Professional Documents
Culture Documents
(9th Edition)
Chapter 11
Analysis of Variance
Chap 11-2
© 2004 Prentice-Hall, Inc.
Chapter Topics (continued)
Chap 11-3
© 2004 Prentice-Hall, Inc.
General Experimental Setting
Investigator Controls One or More
Independent Variables
Called treatment variables or factors
Each treatment factor contains two or more groups
(or levels)
Observe Effects on Dependent Variable
Response to groups (or levels) of independent
variable
Experimental Design: The Plan Used to Test
Hypothesis
Chap 11-4
© 2004 Prentice-Hall, Inc.
Completely Randomized Design
Experimental Units (Subjects) are Assigned
Randomly to Groups
Subjects are assumed to be homogeneous
Only One Factor or Independent Variable
With 2 or more groups (or levels)
Analyzed by One-Way Analysis of Variance
(ANOVA)
Chap 11-5
© 2004 Prentice-Hall, Inc.
Completely Randomized Design
Example
Units
21 hrs 17 hrs 31 hrs
Dependent
Variable 27 hrs 25 hrs 28 hrs
(Response)
29 hrs 20 hrs 22 hrs
Chap 11-6
© 2004 Prentice-Hall, Inc.
One-Way Analysis of Variance
F Test
Evaluate the Difference Among the Mean
Responses of 2 or More (c ) Populations
E.g., Several types of tires, oven temperature settings
Assumptions
Samples are randomly and independently drawn
This condition must be met
Populations are normally distributed
F Test is robust to moderate departure from normality
Chap 11-7
© 2004 Prentice-Hall, Inc.
Why ANOVA?
Could Compare the Means One Pair at a Time
Using t Test for Difference of Means
Each t Test Contains Type I Error
The Total Type I Error with k Pairs of Means
is 1- (1 - ) k
E.g., If there are 5 means and use = .05
Must perform 10 comparisons
Type I Error is 1 – (.95) 10 = .40
40% of the time you will reject the null hypothesis
of equal means in favor of the alternative when the
null hypothesis is true!
Chap 11-8
© 2004 Prentice-Hall, Inc.
Hypotheses of One-Way ANOVA
H 0 : 1 2 c
All population means are equal
No treatment effect (no variation in means among
groups)
H1 : Not all j are the same
At least one population mean is different (others
may be the same!)
There is a treatment effect
Does not mean that all population means are
different
Chap 11-9
© 2004 Prentice-Hall, Inc.
One-Way ANOVA
(No Treatment Effect)
H 0 : 1 2 c
H1 : Not all j are the same
The Null
Hypothesis is
True
1 2 3
Chap 11-10
© 2004 Prentice-Hall, Inc.
One-Way ANOVA
(Treatment Effect Present)
H 0 : 1 2 c
H1 : Not all j are the same The Null
Hypothesis is
NOT True
1 2 3 1 2 3
Chap 11-11
© 2004 Prentice-Hall, Inc.
One-Way ANOVA
(Partition of Total Variation)
Total Variation SST
SST ( X ij X ) 2
j 1 i 1
X
j 1 i 1
ij
X
2 2 2
SST X 11 X 21 X X nc c X
Response, X
Chap 11-15
© 2004 Prentice-Hall, Inc.
Among-Group Variation(continued)
2 2 2
SSA n1 X 1 X n2 X 2 X nc X c X
Response, X
X2 X3
X
X1
SSW X 11 X 1 X 21 X 1 X nc c X c
2 2 2
Response, X
X2 X3
X
X1
j
Chap 11-19
© 2004 Prentice-Hall, Inc.
One-Way ANOVA
F Test Statistic
Test Statistic
MSA
F
MSW
MSA is mean squares among
MSW is mean squares within
Degrees of Freedom
df1 c 1
df 2 n c
Chap 11-20
© 2004 Prentice-Hall, Inc.
One-Way ANOVA
Summary Table
Degrees Mean
Source of Sum of F
of Squares
Variation Squares Statistic
Freedom (Variance)
Among MSA =
c–1 SSA MSA/MSW
(Factor) SSA/(c – 1 )
Within MSW =
n–c SSW
(Error) SSW/(n – c )
SST =
Total n–1
SSA + SSW
Chap 11-21
© 2004 Prentice-Hall, Inc.
Features of One-Way ANOVA
F Statistic
The F Statistic is the Ratio of the Among
Estimate of Variance and the Within Estimate
of Variance
The ratio must always be positive
df1 = c -1 will typically be small
df2 = n - c will typically be large
The Ratio Should Be Close to 1 if the Null is
True
Chap 11-22
© 2004 Prentice-Hall, Inc.
Features of One-Way ANOVA
F Statistic (continued)
Chap 11-23
© 2004 Prentice-Hall, Inc.
One-Way ANOVA F Test
Example
Chap 11-24
© 2004 Prentice-Hall, Inc.
One-Way ANOVA Example:
Scatter Diagram
Machine1 Machine2 27
Time in Seconds
Machine3 •
26
25.40 23.40 20.00 ••
26.31 21.80 22.20 25 X1
24 •
24.10 23.50 19.75 • ••
23.74 22.75 20.60 23 •
X2 • X
25.10 21.60 20.40 22 ••
21
X 1 24.93 X 2 22.61 •• X3
20 ••
X 3 20.59 X 22.71 19
Chap 11-25
© 2004 Prentice-Hall, Inc.
One-Way ANOVA Example
Computations
Machine1 Machine2 Machine3 X 1 24.93 nj 5
25.40 23.40 20.00
26.31 21.80 22.20 X 2 22.61 c3
24.10 23.50 19.75 X 3 20.59 n 15
23.74 22.75 20.60
25.10 21.60 20.40 X 22.71
Degrees Mean
Source of Sum of F
of Squares
Variation Squares Statistic
Freedom (Variance)
Among MSA/MSW
3-1=2 47.1640 23.5820
(Factor) =25.60
Within
15-3=12 11.0532 .9211
(Error)
Total 15-1=14 58.2172
Chap 11-27
© 2004 Prentice-Hall, Inc.
One-Way ANOVA Example
Solution
Test Statistic:
H0: 1 = 2 = 3
H1: Not All Equal MSA 23.5820
= .05 F 25.6
df1= 2 df2 = 12
MSW .9211
Decision:
Critical Value(s):
Reject at = 0.05.
= 0.05 Conclusion:
There is evidence that at
least one j differs from
0 3.89 F the rest.
Chap 11-28
© 2004 Prentice-Hall, Inc.
Solution in Excel
Use Tools | Data Analysis | ANOVA: Single
Factor
Excel Worksheet that Performs the One-Factor
ANOVA of the Example
Microsoft Excel
Worksheet
Chap 11-29
© 2004 Prentice-Hall, Inc.
The Tukey-Kramer Procedure
Tells which Population Means are Significantly
Different
E.g., 1 = 2 3 f(X)
2 groups whose means
may be significantly
different 1= 2 3 X
Post Hoc (A Posteriori) Procedure
Done after rejection of equal means in ANOVA
Pairwise Comparisons
Compare absolute mean differences with critical range
Chap 11-30
© 2004 Prentice-Hall, Inc.
The Tukey-Kramer Procedure:
Example
1. Compute absolute mean
Machine1 Machine2 Machine3 differences:
25.40 23.40 20.00
X 1 X 2 24.93 22.61 2.32
26.31 21.80 22.20
24.10 23.50 19.75 X 1 X 3 24.93 20.59 4.34
23.74 22.75 20.60
25.10 21.60 20.40 X 2 X 3 22.61 20.59 2.02
2. Compute critical range:
MSW 1 1
Critical Range QU ( c ,n c ) 1.618
2 nj nj'
3. All of the absolute mean differences are greater than
the critical range. There is a significant difference between
each pair of means at the 5% level of significance.
Chap 11-31
© 2004 Prentice-Hall, Inc.
Solution in PHStat
Use PHStat | c-Sample Tests | Tukey-Kramer
Procedure …
Excel Worksheet that Performs the Tukey-
Kramer Procedure for the Previous Example
Microsoft Excel
Worksheet
Chap 11-32
© 2004 Prentice-Hall, Inc.
Levene’s Test for
Homogeneity of Variance
The Null Hypothesis
H 0 : 12 22 c2
The c population variances are all equal
The Alternative Hypothesis
H : Not all
j are equal ( j 1, 2, , c )
2
1
Not all the c population variances are equal
Chap 11-33
© 2004 Prentice-Hall, Inc.
Levene’s Test for Homogeneity
of Variance: Procedure
Chap 11-34
© 2004 Prentice-Hall, Inc.
Levene’s Test for Homogeneity
of Variances: Example
Chap 11-35
© 2004 Prentice-Hall, Inc.
Levene’s Test:
Absolute Difference from the Median
Chap 11-36
© 2004 Prentice-Hall, Inc.
Summary Table
SUMMARY
Groups Count Sum Average Variance
Machine1 5 3.87 0.774 0.35208
Machine2 5 3.5 0.7 0.19
Machine3 5 3.05 0.61 0.5005
ANOVA
Source of Variation SS df MS F P-value F crit
Between Groups 0.067453 2 0.033727 0.097048 0.908218 3.88529
Within Groups 4.17032 12 0.347527
Total 4.237773 14
Chap 11-37
© 2004 Prentice-Hall, Inc.
Levene’s Test Example:
Solution
Chap 11-39
© 2004 Prentice-Hall, Inc.
Randomized Blocked Design
(Example)
Factor (Training Method)
Factor Levels
(Groups)
Blocked
Experiment
Units
21 hrs 17 hrs 31 hrs
Dependent
Variable 27 hrs 25 hrs 28 hrs
(Response)
29 hrs 20 hrs 22 hrs
Chap 11-40
© 2004 Prentice-Hall, Inc.
Randomized Block Design
(Partition of Total Variation)
Variation Due Commonly referred to as:
to Group Sum of Squares Among
+
Variation Commonly referred to as:
Variation Among Sum of Squares Among
Among All Blocks Blocks
Observations = SSBL
SST +
Variation Due Commonly referred to as:
to Random Sum of Squares Error
SSW Unexplained
Chap 11-41
© 2004 Prentice-Hall, Inc.
Total Variation
c r 2
SST X ij X
j 1 i 1
c 2
SSA r X j X
j 1
r
X ij
X j i 1
(treatment group means)
r
df c 1
SSA
MSA
c 1
Chap 11-43
© 2004 Prentice-Hall, Inc.
Among-Block Variation
r 2
SSBL c X i X
i 1
c
X
j 1
ij
X i (block means)
c
df r 1
SSBL
MSBL
r 1
Chap 11-44
© 2004 Prentice-Hall, Inc.
Random Error
c r 2
SSE X ij X i X j X
j 1 i 1
df r 1 c 1
SSE
MSE
r 1 c 1
Chap 11-45
© 2004 Prentice-Hall, Inc.
Randomized Block F Test for
Differences in c Means
H 0 : 1 2 c
No treatment effect
H1 : Not all j are equal
Test Statistic Reject
MSA
F
MSE
Degrees of Freedom
df1 c 1 0 FU F
df 2 r 1 c 1
Chap 11-46
© 2004 Prentice-Hall, Inc.
Summary Table
MSE =
Error (r – 1) c – 1) SSE
SSE/[(r – 1)(c– 1)]
Total rc – 1 SST
Chap 11-47
© 2004 Prentice-Hall, Inc.
Randomized Block Design:
Example
As production manager, you want
to see if 3 filling machines have Machine1 Machine2
different mean filling times. You Machine3
assign 15 workers with varied 25.40 23.40 20.00
experience into 5 groups of 3
26.31 21.80 22.20
based on similarity of their
experience, and assigned each 24.10 23.50 19.75
group of 3 workers with similar 23.74 22.75 20.60
experience to the machines. At 25.10 21.60 20.40
the .05 significance level, is there
a difference in mean filling times?
Chap 11-48
© 2004 Prentice-Hall, Inc.
Randomized Block Design
Example Computation
Machine1 Machine2 Machine3 X 1 24.93 r 5
25.40 23.40 20.00
26.31 21.80 22.20 X 2 22.61 c3
24.10 23.50 19.75 X 3 20.59 n 15
23.74 22.75 20.60
25.10 21.60 20.40 X 22.71
SSA 5 24.93 22.71 22.61 22.71 20.59 22.71
2 2 2
47.164
SSE 8.4025
MSA SSA /(c -1) 47.16 / 2 23.5820
MSE SSE / (r -1) c 1 8.4025 / 8 1.0503
Chap 11-49
© 2004 Prentice-Hall, Inc.
Randomized Block Design
Example: Summary Table
SSE= MSE =
Error
8.4025 1.0503
SST=
Total 14
58.2172
Chap 11-50
© 2004 Prentice-Hall, Inc.
Randomized Block Design
Example: Solution
H0: 1 = 2 = 3 Test Statistic:
H1: Not All Equal MSA 23.582
= .05 F 22.45
df1= 2 df2 = 8 MSE 1.0503
Microsoft Excel
Worksheet
Chap 11-52
© 2004 Prentice-Hall, Inc.
The Tukey Procedure
Similar to the Tukey Procedure for the
Completely Randomized Design Case
Critical Range
MSE
Critical Range QU ( c , r 1 c 1 )
r
Chap 11-53
© 2004 Prentice-Hall, Inc.
The Tukey Procedure: Example
1. Compute absolute mean
Machine1 Machine2 Machine3 differences:
25.40 23.40 20.00
X 1 X 2 24.93 22.61 2.32
26.31 21.80 22.20
24.10 23.50 19.75 X 1 X 3 24.93 20.59 4.34
23.74 22.75 20.60
25.10 21.60 20.40 X 2 X 3 22.61 20.59 2.02
2. Compute critical range:
MSE 1.0503
Critical Range QU ( c , r 1 c 1 ) 4.04 1.8516
r 5
3. All of the absolute mean differences are greater. There
is a significance difference between each pair of means at
5% level of significance.
Chap 11-54
© 2004 Prentice-Hall, Inc.
Two-Way ANOVA
Examines the Effect of:
Two factors on the dependent variable
E.g., Percent carbonation and line speed on soft
drink bottling process
Interaction between the different levels of these
two factors
E.g., Does the effect of one particular percentage of
carbonation depend on which level the line speed is
set?
Chap 11-55
© 2004 Prentice-Hall, Inc.
Two-Way ANOVA (continued)
Assumptions
Normality
Populations are normally distributed
Homogeneity of Variance
Populations have equal variances
Independence of Errors
Independent random samples are drawn
Chap 11-56
© 2004 Prentice-Hall, Inc.
Two-Way ANOVA
Total Variation Partitioning
Chap 11-57
© 2004 Prentice-Hall, Inc.
Two-Way ANOVA
Total Variation Partitioning
Chap 11-58
© 2004 Prentice-Hall, Inc.
Total Variation
r c n' 2
SST X ijk X
i 1 j 1 k 1
X
i 1 j 1 k 1
ijk X
i 1 j 1 k 1
ijk
X '
rcn n
the overall or grand mean Chap 11-59
© 2004 Prentice-Hall, Inc.
Factor A Variation
r 2
SSA cn X i X
'
i 1
Chap 11-60
© 2004 Prentice-Hall, Inc.
Factor B Variation
c 2
SSB rn X j X
'
j 1
Chap 11-61
© 2004 Prentice-Hall, Inc.
Interaction Variation
r c 2
SSAB n X ij X i X j X
'
i 1 j 1
Chap 11-62
© 2004 Prentice-Hall, Inc.
Random Error
2
n'
r c
SSE X ijk X ij
i 1 j 1 k 1
Chap 11-63
© 2004 Prentice-Hall, Inc.
Two-Way ANOVA:
The F Test Statistic
H0: 1 ..= 2 .. = ••• = r .. F Test for Factor A Main Effect
MSA SSA Reject if
F MSA
MSE r 1 F > FU
H1: Not all i .. are equal
H0: 1. = .2. = ••• = c. F Test for Factor B Main Effect
MSB SSB Reject if
F MSB
H1: Not all .j. are equal MSE c 1 F > FU
AB MSAB = MSAB/
(r – 1)(c – 1) SSAB
(Interaction) SSAB/ [(r – 1)(c – 1)] MSE
MSE =
Error rc n – 1)
’
SSE
SSE/[rc n’ – 1)]
Total rc n’ – 1 SST
Chap 11-65
© 2004 Prentice-Hall, Inc.
Features of Two-Way ANOVA
F Test
Degrees of Freedom Always Add Up
rcn’-1=rc(n’-1)+(c-1)+(r-1)+(c-1)(r-1)
Total=Error+Column+Row+Interaction
The Denominator of the F Test is Always the
Same but the Numerator is Different
The Sums of Squares Always Add Up
Total=Error+Column+Row+Interaction
Chap 11-66
© 2004 Prentice-Hall, Inc.
Kruskal-Wallis Rank Test
Used to Analyze Completely Randomized
Experimental Designs
Extension of Wilcoxon Rank Sum Test
Tests the equality of more than 2 population
medians
Distribution-Free Test Procedure
Use 2 Distribution to Approximate if Each
Sample Group Size nj > 5
df = c – 1
Chap 11-67
© 2004 Prentice-Hall, Inc.
Kruskal-Wallis Rank Test
Assumptions
Independent random samples are drawn
Continuous dependent variable
Data may be ranked both within and among
samples
Populations have same variability
Populations have same shape
Robust with Regard to Last 2 Conditions
Use F test in completely randomized designs and
when the more stringent assumptions hold
Chap 11-68
© 2004 Prentice-Hall, Inc.
Kruskal-Wallis Rank Test
Procedure
Obtain Ranks
In event of tie, each of the tied values gets their
average rank
Add the Ranks for Data from Each of the c
12 c T2
H
Groups j
3(n 1)
n(ton obtain
Square
1) j 1 Tn j2
j
n n1 n2 nc
Chap 11-69
© 2004 Prentice-Hall, Inc.
Kruskal-Wallis Rank Test
Procedure (continued)
Chap 11-71
© 2004 Prentice-Hall, Inc.
Kruskal-Wallis Rank Test:
Example
Machine1 Machine2
As production manager, you Machine3
want to see if 3 filling machines 25.40 23.40 20.00
have different median filling 26.31 21.80 22.20
times. You assign 15 similarly
24.10 23.50 19.75
trained & experienced workers,
23.74 22.75 20.60
5 per machine, to the
25.10 21.60 20.40
machines. At the .05
significance level, is there a
difference in median filling
times?
Chap 11-72
© 2004 Prentice-Hall, Inc.
Example Solution: Step 1
Obtaining a Ranking
Raw Data Ranks
Machine1 Machine2 Machine1 Machine2
Machine3 Machine3
25.40 23.40 20.00 14 9 2
26.31 21.80 22.20 15 6 7
24.10 23.50 19.75 12 10 1
23.74 22.75 20.60 11 8 4
25.10 21.60 20.40 13
65 5
38 3
17
Chap 11-73
© 2004 Prentice-Hall, Inc.
Example Solution: Step 2
Test Statistic Computation
T 2
12 c j
H 3(n 1)
n(n 1) j 1 n
j
12 652 382 17 2
3(15 1)
15(15 1) 5 5 5
11.58
Chap 11-74
© 2004 Prentice-Hall, Inc.
Kruskal-Wallis Test Example
Solution
H0: M1 = M2 = M3 Test Statistic:
H1: Not all equal H = 11.58
= .05
df = c - 1 = 3 - 1 = 2 Decision:
Critical Value(s): Reject at = .05.
Conclusion:
= .05
There is evidence that
population medians are
not all equal.
0 5.991
Chap 11-75
© 2004 Prentice-Hall, Inc.
Kruskal-Wallis Test in PHStat
PHStat | c-Sample Tests | Kruskal-Wallis Rank
Sum Test …
Example Solution in Excel Spreadsheet
Microsoft Excel
Worksheet
Chap 11-76
© 2004 Prentice-Hall, Inc.
Friedman Rank Test
Used to Analyze Randomized Block
Experimental Designs
Tests the equality of more than 2 population
medians
Distribution-Free Test Procedure
Use 2 Distribution to Approximate if the
Number of Blocks r > 5
df = c – 1
Chap 11-77
© 2004 Prentice-Hall, Inc.
Friedman Rank Test
Assumptions
The r blocks are independent
The random variable is continuous
The data constitute at least an ordinal scale of
measurement
No interaction between the r blocks and the c
treatment levels
The c populations have the same variability
The c populations have the same shape
Chap 11-78
© 2004 Prentice-Hall, Inc.
Friedman Rank Test:
Procedure
Replace the c observations by their ranks in
each of the r blocks; assign average rank
for ties
c
12
Test statistic: FR
rc c 1 j 1
R j 3r c 1
2
Chap 11-79
© 2004 Prentice-Hall, Inc.
Friedman Rank Test: Example
As production manager, you want
to see if 3 filling machines have Machine1 Machine2
different median filling times. You
Machine3
assign 15 workers with varied
experience into 5 groups of 3 25.40 23.40 20.00
based on similarity of their 26.31 21.80 22.20
experience, and assigned each
group of 3 workers with similar
24.10 23.50 19.75
experience to the machines. At 23.74 22.75 20.60
the .05 significance level, is there 25.10 21.60 20.40
a difference in median filling
times?
Chap 11-80
© 2004 Prentice-Hall, Inc.
Friedman Rank Test:
Computation Table
Timing Rank
Machine 1 Machine 2 Machine 3 Machine 1 Machine 2 Machine 3
25.4 23.4 20 3 2 1
26.31 21.8 22.2 3 1 2
24.1 23.5 19.75 3 2 1
23.74 22.75 20.6 3 2 1
25.1 21.6 20.4 3 2 1
R. j 15 9 6
R.2j 225 81 36
c
12
FR
rc c 1
j 3r c 1
R 2
j 1
12
342 3 5 4 8.4
5 3 4 Chap 11-81
© 2004 Prentice-Hall, Inc.
Friedman Rank Test Example
Solution
H0: M1 = M2 = M3 Test Statistic:
H1: Not all equal FR = 8.4
= .05
df = c - 1 = 3 - 1 = 2 Decision:
Critical Value: Reject at = .05
Conclusion:
= .05
There is evidence that
population medians are
not all equal.
0 5.991
Chap 11-82
© 2004 Prentice-Hall, Inc.
Chapter Summary
Described the Completely Randomized
Design: One-Way Analysis of Variance
ANOVA Assumptions
F Test for Differences in More than Two Means
The Tukey-Kramer Procedure
Levene’s Test for Homogeneity of Variance
Discussed the Randomized Block Design
F Test for the Difference in More than Two Means
The Tukey Procedure
Chap 11-83
© 2004 Prentice-Hall, Inc.
Chapter Summary (continued)